Author: Robert Pawlak
Date: 12:50:07 10/23/00
A number of factors have lead to me to recently take another look at the state of playing programs. Sure, we've got stronger engines than ever, but it seems as though no one has yet to develop a really good method for handicapping program strength. Some of you may remember a recent thread here that discussed the use of weaker Winboard engines for play against humans. So there are some people that find it difficult to get a good game out of their commercial software. By "good" I mean a game in which the poor software owner does not get blown off the board in 30 moves or less. I've made a partial list of the current crop of commercial engines, along with what I believe are issues with the ability to handicap engine play for each: Chessmaster 6000/7000 - Blunders away material early on in the game, then plays like the All Mighty until it's opponent is crushed. Fritz 6 - Still plays too quickly on the ELO handicaping levels. Tends to leave it's king in the middle of the board in friend mode. The Sparring setting can sometimes be interesting, howver, it seems to play very well in the absence of tactical threats... Rebel - ELO handicapping used to work pretty well. Indeed I thought Rebel was the best in this respect. But now I have trouble due to a faster machine/different CPU. In general, it plays a very strong game, even at the lower ELO levels (and with few, if any, tactical mistakes). There are plenty of other programs (i.e. Shredder) that do not support ELO handicapping at all. Clearly, program strength sells. And it seems to be where everyone's interest is. But I really think a bit more attention needs to be paid to satisfying those of us that are not Masters, and really want to use a playing program for it's intended purpose! Can someone tell me why there is not more emphasis placed on this aspect of engine design? The obvious answer is that it is not very glamorous, and therefore cannot be used to sell a program. But let's set this excuse aside for a momement. Is it mostly that there is no perceived need, or is it a problem with calibration, testing, or handicapping methods? Is there a solution? Does anyone even care? Sometimes it seems as though most people reading this board are playing engine matches, and not really playig against the programs themselves. And that's fine, but is it representative of what the typical player expects from his/her software? Bob Pawlak
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.